import json import os from os.path import join as p_join from tqdm import tqdm from time import time import hdbscan import numpy as np import matplotlib.pyplot as plt import seaborn as sns from sklearn.manifold import TSNE import pandas as pd from datasets import load_dataset from model_meta_voice import MetaVoiceEmbedding from model_pyannote_embedding import PyannoteEmbedding from model_w2v_bert import W2VBERTEmbedding from model_clap import CLAPEmbedding, CLAPGeneralEmbedding from model_xls import Wav2VecEmbedding, XLSR300MEmbedding, XLSR1BEmbedding, XLSR2BEmbedding from model_hubert import HuBERTBaseEmbedding, HuBERTLargeEmbedding, HuBERTXLEmbedding def get_embedding(model_class, model_name: str, dataset_name: str, data_split: str): dataset = load_dataset(dataset_name, split=data_split) file_path = p_join("experiment_cache", "embeddings", f"{model_name}.{os.path.basename(dataset_name)}.json") os.makedirs(os.path.dirname(file_path), exist_ok=True) if os.path.exists(file_path): return model = model_class() embeddings = [] for i in tqdm(dataset, total=len(dataset)): start = time() v = model.get_speaker_embedding(i["audio"]["array"], i["audio"]["sampling_rate"]) tmp = { "model": model_name, "embedding": v.tolist(), "sampling_rate": i["audio"]["sampling_rate"], "process_time": time() - start, "dataset_name": os.path.basename(dataset_name) } tmp.update({k: v for k, v in i.items() if k != "audio"}) embeddings.append(tmp) with open(file_path, "w") as f: f.write("\n".join([json.dumps(i) for i in embeddings])) def cluster_embedding(model_name, dataset_name, label_name: str): file_path_embedding = p_join("experiment_cache", "embeddings", f"{model_name}.{os.path.basename(dataset_name)}.json") file_path_cluster = p_join("experiment_cache", "cluster", f"{model_name}.{os.path.basename(dataset_name)}.{label_name}.csv") if not os.path.exists(file_path_cluster): print('CLUSTERING') os.makedirs(os.path.dirname(file_path_cluster), exist_ok=True) assert os.path.exists(file_path_embedding) with open(file_path_embedding) as f: data = [json.loads(i) for i in f.readlines()] clusterer = hdbscan.HDBSCAN() embeddings = [i["embedding"] for i in data] keys = [i["id"] for i in data] clusterer.fit(np.stack(embeddings)) # data x dimension print(f'{clusterer.labels_.max()} clusters found from {len(data)} data points') print(f"generating report for {label_name}") label = [i[label_name] for i in data] cluster_info = [ {"id": k, "cluster": c, f"label.{label_name}": l} for c, k, l in zip(clusterer.labels_, keys, label) if c != -1 ] cluster_df = pd.DataFrame(cluster_info) cluster_df.to_csv(file_path_cluster, index=False) file_path_tsne = p_join("experiment_cache", "tsne", f"{model_name}.{os.path.basename(dataset_name)}.{label_name}.npy") if not os.path.exists(file_path_tsne): os.makedirs(os.path.dirname(file_path_tsne), exist_ok=True) print('DIMENSION REDUCTION') assert os.path.exists(file_path_embedding) with open(file_path_embedding) as f: data = np.stack([json.loads(i)['embedding'] for i in f.readlines()]) # data x dimension print(f'Dimension reduction: {data.shape}') embedding_2d = TSNE(n_components=2, random_state=0).fit_transform(data) np.save(file_path_tsne, embedding_2d) embedding_2d = np.load(file_path_tsne) print('PLOT') figure_path = p_join("experiment_cache", "figure", f"2d.latent_space.{model_name}.{os.path.basename(dataset_name)}.{label_name}.png") os.makedirs(os.path.dirname(figure_path), exist_ok=True) with open(file_path_embedding) as f: label = np.stack([json.loads(i)[label_name] for i in f.readlines()]) # data x dimension label_type = sorted(list(set(label))) label2id = {v: n for n, v in enumerate(label_type)} plt.figure() scatter = plt.scatter( embedding_2d[:, 0], embedding_2d[:, 1], s=8, c=[label2id[i] for i in label], cmap=sns.color_palette('Spectral', len(label_type), as_cmap=True) ) plt.gca().set_aspect('equal', 'datalim') plt.legend(handles=scatter.legend_elements(num=len(label_type))[0], labels=label_type, bbox_to_anchor=(1.04, 1), borderaxespad=0, loc='upper left', ncol=3 if len(label2id) > 12 else 1) plt.savefig(figure_path, bbox_inches='tight', dpi=600) def analyze_embedding(model_name: str, dataset_name: str, n_shot: int = 5, n_cross_validation: int = 5): file_path = p_join("experiment_cache", "embeddings", f"{model_name}.{os.path.basename(dataset_name)}.json") assert os.path.exists(file_path) with open(file_path) as f: embeddings = [json.loads(i) for i in f.readlines()] df = pd.DataFrame(embeddings) process_time = df["process_time"].mean() df.groupby("speaker_ido") sorted(df["speaker_id"].unique()) if __name__ == '__main__': # get_embedding(MetaVoiceEmbedding, "meta_voice_se", "asahi417/voxceleb1-test-split", "test") # get_embedding(PyannoteEmbedding, "pyannote_se", "asahi417/voxceleb1-test-split", "test") # get_embedding(W2VBERTEmbedding, "w2v_bert_se", "asahi417/voxceleb1-test-split", "test") # get_embedding(CLAPEmbedding, "clap_se", "asahi417/voxceleb1-test-split", "test") # get_embedding(CLAPGeneralEmbedding, "clap_general_se", "asahi417/voxceleb1-test-split", "test") # get_embedding(HuBERTBaseEmbedding, "hubert_base_se", "asahi417/voxceleb1-test-split", "test") # get_embedding(HuBERTLargeEmbedding, "hubert_large_se", "asahi417/voxceleb1-test-split", "test") # get_embedding(HuBERTXLEmbedding, "hubert_xl_se", "asahi417/voxceleb1-test-split", "test") get_embedding(Wav2VecEmbedding, "wav2vec_se", "asahi417/voxceleb1-test-split", "test") get_embedding(XLSR300MEmbedding, "xlsr_300m_se", "asahi417/voxceleb1-test-split", "test") get_embedding(XLSR1BEmbedding, "xlsr_1b_se", "asahi417/voxceleb1-test-split", "test") get_embedding(XLSR2BEmbedding, "xlsr_2b_se", "asahi417/voxceleb1-test-split", "test") # get_embedding(MetaVoiceEmbedding, "meta_voice_se", "ylacombe/expresso", "train") # get_embedding(PyannoteEmbedding, "pyannote_se", "ylacombe/expresso", "train") # get_embedding(W2VBERTEmbedding, "w2v_bert_se", "ylacombe/expresso", "train") # get_embedding(CLAPEmbedding, "clap_se", "ylacombe/expresso", "train") # get_embedding(CLAPGeneralEmbedding, "clap_general_se", "ylacombe/expresso", "train") # get_embedding(HuBERTBaseEmbedding, "hubert_base_se", "ylacombe/expresso", "train") # get_embedding(HuBERTLargeEmbedding, "hubert_large_se", "ylacombe/expresso", "train") # get_embedding(HuBERTXLEmbedding, "hubert_xl_se", "ylacombe/expresso", "train") get_embedding(Wav2VecEmbedding, "wav2vec_se", "ylacombe/expresso", "train") get_embedding(XLSR300MEmbedding, "xlsr_300m_se", "ylacombe/expresso", "train") get_embedding(XLSR1BEmbedding, "xlsr_1b_se", "ylacombe/expresso", "train") get_embedding(XLSR2BEmbedding, "xlsr_2b_se", "ylacombe/expresso", "train") # cluster_embedding("meta_voice_se", "asahi417/voxceleb1-test-split", "speaker_id") # cluster_embedding("pyannote_se", "asahi417/voxceleb1-test-split", "speaker_id") # cluster_embedding("w2v_bert_se", "asahi417/voxceleb1-test-split", "speaker_id") # cluster_embedding("clap_se", "asahi417/voxceleb1-test-split", "speaker_id") # cluster_embedding("clap_general_se", "asahi417/voxceleb1-test-split", "speaker_id") # cluster_embedding("hubert_base_se", "asahi417/voxceleb1-test-split", "speaker_id") # cluster_embedding("hubert_large_se", "asahi417/voxceleb1-test-split", "speaker_id") # cluster_embedding("hubert_xl_se", "asahi417/voxceleb1-test-split", "speaker_id") cluster_embedding("wav2vec_se", "asahi417/voxceleb1-test-split", "speaker_id") cluster_embedding("xlsr_300m_se", "asahi417/voxceleb1-test-split", "speaker_id") cluster_embedding("xlsr_1b_se", "asahi417/voxceleb1-test-split", "speaker_id") cluster_embedding("xlsr_2b_se", "asahi417/voxceleb1-test-split", "speaker_id") # cluster_embedding("meta_voice_se", "ylacombe/expresso", "speaker_id") # cluster_embedding("pyannote_se", "ylacombe/expresso", "speaker_id") # cluster_embedding("w2v_bert_se", "ylacombe/expresso", "speaker_id") # cluster_embedding("clap_se", "ylacombe/expresso", "speaker_id") # cluster_embedding("clap_general_se", "ylacombe/expresso", "speaker_id") # cluster_embedding("hubert_base_se", "ylacombe/expresso", "speaker_id") # cluster_embedding("hubert_large_se", "ylacombe/expresso", "speaker_id") # cluster_embedding("hubert_xl_se", "ylacombe/expresso", "speaker_id") cluster_embedding("wav2vec_se", "ylacombe/expresso", "speaker_id") cluster_embedding("xlsr_300m_se", "ylacombe/expresso", "speaker_id") cluster_embedding("xlsr_1b_se", "ylacombe/expresso", "speaker_id") cluster_embedding("xlsr_2b_se", "ylacombe/expresso", "speaker_id") # cluster_embedding("meta_voice_se", "ylacombe/expresso", "style") # cluster_embedding("pyannote_se", "ylacombe/expresso", "style") # cluster_embedding("w2v_bert_se", "ylacombe/expresso", "style") # cluster_embedding("clap_se", "ylacombe/expresso", "style") # cluster_embedding("clap_general_se", "ylacombe/expresso", "style") # cluster_embedding("hubert_base_se", "ylacombe/expresso", "style") # cluster_embedding("hubert_large_se", "ylacombe/expresso", "style") # cluster_embedding("hubert_xl_se", "ylacombe/expresso", "style") cluster_embedding("wav2vec_se", "ylacombe/expresso", "style") cluster_embedding("xlsr_300m_se", "ylacombe/expresso", "style") cluster_embedding("xlsr_1b_se", "ylacombe/expresso", "style") cluster_embedding("xlsr_2b_se", "ylacombe/expresso", "style")